Anxiety Level Prediction Using Physiological Data From Wrist-worn Wearable Device

Anxiety disorders form a group of mental disorders which are quite common and have been shown to affect millions of people around the world. There are a number of traditional approaches to assessment of anxiety, and typically they are based on the subject’s questionnaires. This report outlines a new solution that is based on the accurate calculation of anxiety levels based on physiological parameters collected by wrist-worn devices. Here, the following study is proposed to employ a broad dataset that includes physiological data with citizens wearing wrist-worn gadgets. Five ML algorithms are employed for anxiety level prediction: K-Nearest Neighbors, Support Vector Machines, Decision trees, Random forests, Histogram Based Gradient Boosting etc. As for the validation of these models, cross-validation of the k-fold type is used to increase the model’s reliability and to make it as universal as possible. As it will be shown next, the proposed approach has promising applications for refining the evaluation and treatment of anxiety. As such, the present research aims to extend the existing literature on self-report measures for anxiety levels and offer an actual physiological perspective by using only basic physiological data and applying the sophisticated methods of AI. The results of this study might positively influence the creation of online surveillance programs and appropriate strategies for proactive treatment of anxiety disorders that would improve the subjects’ quality of life. This study helps to advance the area of affective computing to uncover how analysis of physiological data, machine learning algorithms, and mental health can all be connected. Among the possibilities, further developments of the current models can be mentioned as well as the promoted cooperation between disciplines and the investigation of the bodily expressions of anxiety.